nikolamilosevic
commited on
Update README.md
Browse filesAdded some description on dataset
README.md
CHANGED
@@ -1,28 +1,50 @@
|
|
1 |
-
---
|
2 |
-
license: agpl-3.0
|
3 |
-
configs:
|
4 |
-
- config_name: default
|
5 |
-
data_files:
|
6 |
-
- split: train
|
7 |
-
path: data/train-*
|
8 |
-
- split: val
|
9 |
-
path: data/val-*
|
10 |
-
- split: test
|
11 |
-
path: data/test-*
|
12 |
-
dataset_info:
|
13 |
-
features:
|
14 |
-
- name: input
|
15 |
-
dtype: string
|
16 |
-
splits:
|
17 |
-
- name: train
|
18 |
-
num_bytes: 136602851.95652175
|
19 |
-
num_examples: 7260
|
20 |
-
- name: val
|
21 |
-
num_bytes: 17065948.584650856
|
22 |
-
num_examples: 907
|
23 |
-
- name: test
|
24 |
-
num_bytes: 17084764.40447958
|
25 |
-
num_examples: 908
|
26 |
-
download_size: 82888007
|
27 |
-
dataset_size: 170753564.9456522
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: agpl-3.0
|
3 |
+
configs:
|
4 |
+
- config_name: default
|
5 |
+
data_files:
|
6 |
+
- split: train
|
7 |
+
path: data/train-*
|
8 |
+
- split: val
|
9 |
+
path: data/val-*
|
10 |
+
- split: test
|
11 |
+
path: data/test-*
|
12 |
+
dataset_info:
|
13 |
+
features:
|
14 |
+
- name: input
|
15 |
+
dtype: string
|
16 |
+
splits:
|
17 |
+
- name: train
|
18 |
+
num_bytes: 136602851.95652175
|
19 |
+
num_examples: 7260
|
20 |
+
- name: val
|
21 |
+
num_bytes: 17065948.584650856
|
22 |
+
num_examples: 907
|
23 |
+
- name: test
|
24 |
+
num_bytes: 17084764.40447958
|
25 |
+
num_examples: 908
|
26 |
+
download_size: 82888007
|
27 |
+
dataset_size: 170753564.9456522
|
28 |
+
task_categories:
|
29 |
+
- text-generation
|
30 |
+
- question-answering
|
31 |
+
- summarization
|
32 |
+
language:
|
33 |
+
- en
|
34 |
+
tags:
|
35 |
+
- biology
|
36 |
+
- biomedicine
|
37 |
+
pretty_name: PubMed Referenced Question Answering Dataset
|
38 |
+
size_categories:
|
39 |
+
- 1K<n<10K
|
40 |
+
---
|
41 |
+
# Dataset description
|
42 |
+
|
43 |
+
The PQAref dataset is a dataset for fine-tuning large language models for referenced question-answering in biomedical domain.
|
44 |
+
|
45 |
+
The dataset contains 3 components:
|
46 |
+
- Instruction - question that is supposed to be answered
|
47 |
+
- Abstracts - set of 10 relevant abstracts retrieved by IR system from PubMed, containing PubMed id, abstract title and the content of the abstract
|
48 |
+
- Answer - expected answer, with references
|
49 |
+
|
50 |
+
The dataset was created semi-automatically, utilizing questions available from PubMedQA dataset.
|